Professional Documents
Culture Documents
Singh Belief As Commitment To The Truth
Singh Belief As Commitment To The Truth
Keshav Singh
Forthcoming in What is Belief? eds. Eric Schwitzgebel and Jonathan Jong
Introduction 1
trying to understand the nature of belief, but it doesn’t take us very far by itself. To give
an account of what belief is from this starting point, we need to investigate in what sense
belief involves regarding a proposition as true, such that this distinguishes it from other
attitudes. 2 For example, we might say that to regard some proposition p as true is to have
an attitude with the content p, where the nature of the attitude is that it represents p as
true. But this still wouldn’t distinguish belief from other, similar attitudes that also
anywhere from the idea that to believe p is to regard it as true, we need to be much more
specific.
true, only in the case of belief does this representation rise to a commitment to the truth
of p. My goal in this essay is to develop and defend this approach to understanding the
to the truth of p. To commit to the truth of p, in the sense I am interested in, is to exercise
1 Thanks to Aliosha Barranco Lopez, Jonathan Jong, Eric Marcus, Ram Neta, Eric Schwitzgebel, Daniel
Wodak, and an anonymous reviewer for comments on an earlier draft of this paper.
2 This is a common point; see, e.g., Velleman (2000) and Shah and Velleman (2005).
1
of truth. This entails that there is a kind of voluntary activity involved in forming and
holding beliefs.
As I’ll argue, my account of belief as commitment to the truth can explain a variety
of features of belief that separate it from other regarding-as-true attitudes. For example,
don’t need any evidence that p is true for it to be appropriate for me to use the assumption
standardly thought that I should do so on the basis of good evidence that p is true. On my
account, the fact that belief is constituted by a commitment to the truth explains the
commitment to the truth operative in my account. In §2, I with the main machinery of my
account on the table, I explain how it captures the unique features of belief that
commitment-based account can help us solve three further puzzles about belief, regarding
doxastic voluntarism, the aim of belief, and Moore’s paradox. I conclude by briefly
1. Commitment
The idea that belief involves commitment to the truth of a proposition has been
endorsed in passing, but rarely explored in any detail. For example, Alston (1996, 24)
writes that “in making statements and holding beliefs we commit ourselves to the
propositional content's being true.” Alston doesn’t elaborate on what he means by this,
2
invoked by epistemologists with related but distinct meanings, so I must explain the
epistemologists have in mind when they say that holding some attitudes commits us to
holding certain other attitudes, whether or not we actually have the latter attitudes. This
is the sense Wedgwood (2002) has in mind when he discusses commitment in relation to
the aim of belief: believing one thing commits one to believing other things. For example,
we might say that believing both p and pq commits one to believing q. According to
Wedgwood, what this means is that, given one’s beliefs that p and pq, it would be
Though it is a useful notion, rational commitment can’t be what is operative in the thought
that belief involves commitment to the truth. Talk of what we are rationally committed to
picks out facts about structural rationality – for example, what I am rationally required
to believe given my other beliefs. While it may follow from the nature of belief that having
some beliefs rationally commits one to having other beliefs, these normative facts can’t
tell us what belief is. Plausibly, what makes me rationally committed to believing q when
I believe both p and pq is that I am committed, in some more basic sense, to the truth
3
of both p and pq, from which the truth of q follows. Only this more basic sense of
that a range of attitudes are commitment-constituted when there is some question the
settling of which amounts to forming that attitude. According to Hieronymi, the sense in
which such attitudes are commitment-constituted is that “if one has the attitude, one is
would be answered by the considerations that bear on the relevant question(s)” (450).
committed to p as true, that is, I am answerable to questions and criticisms that would be
This seems right. It does seem like committing to the truth of p results in my being
answerable to questions and criticisms related to whether p is the case. But this doesn’t
tell us what it is to commit to the truth of p. For it seems like committing to the truth of p
is precisely what makes answerable in the ways Hieronymi describes. Thus, in order to
understand the nature of the commitment that constitutes belief, we must ask: what
happens when I believe p, which makes it the case that I am answerable in these ways?
attitude, I don’t think she tells us what it is to commit to the truth of p. Instead, she gives
4
us a helpful characterization of the relevant kind of commitment: it’s something that
makes one answerable to questions and criticisms related to whether p is the case. But
this answerability is itself a normative relation. For the same reasons discussed above in
may shed light on that attitude, but it doesn’t tell us what the attitude is.
Nevertheless, we’ve made some progress here. We can now ask the question: what
could be going on psychologically when I commit to the truth of p, such that I somehow
make it the case that I am answerable to questions and criticisms related to p? It seems
that in committing to the truth of p, I somehow alter the normative landscape. But how?
“Jacky is committed to Joan” (2014, 148). While he doesn’t say all that much about
volitional commitments, Shpall makes clear that volitional commitments are not
normative facts, but psychological items. They are voluntary exercises of agency that alter
the normative landscape. To say that Jacky is committed to Joan is to say that Jacky has
engaged in some sort of voluntary activity that has altered the normative landscape,
among other things making herself answerable to treating Joan in certain ways.
In her work on normative voluntarism, Chang (2013) sheds light on this volitional
normative powers, the power to confer reason-giving force on something through an act
5
of will” (75, emphasis original). 4 Chang likens this exercise of the will to stipulating the
meaning of the word, another exercise of a power to make something the case just by
willing it be so. It might also be likened to promising, at least on some conceptions thereof.
The important thing is that, for Chang, commitments in the volitional sense are exercises
of a power to make something normative the case. This explains how committing to others
(like Jacky commits to Joan) alters the normative landscape: our commitments to others
are exercises of a normative power to generate special reasons for action that we didn’t
have before.
Now, it might be thought that commitment, as Chang describes it, couldn’t possibly
be involved in belief, for the will is on the side of the practical, and belief is on the side of
the theoretical. Especially given that I want to hold on to the idea that belief is uniquely
connected to evidence, it may seem that I’m playing with fire here. But bear with me.
While the will is sometimes identified with practical reason in particular, I interpret
Chang as having something broader in mind here. In characterizing acts of will, Chang
describes them as “active attitudes” and as “engaging our volition” (2009, 244-245).
There need not be anything essentially practical about the will, so characterized. On the
voluntarist conception of commitment, the important thing is that it involves the exercise
of normative powers to alter the normative landscape, which is an exercise of agency that
To further bring this out, consider how similar the above description of Jacky’s
6
an exercise of agency that alters the normative landscape, making the agent answerable
to certain normative standards. Even if the term ‘will’ is to be reserved for the practical
case, the similarities between these cases suggest that there is a broader kind of exercise
of agency that covers both practical and theoretical cases of commitment in the operative
sense. Indeed, when Hieronymi argues for an identification of the will with practical
reason, she recognizes judgment as analogous capacity on the theoretical side and writes
that “theoretical reasoning involves an activity very much like willing” (2009, 218).
Moreover, she cites Descartes as having “something similar in mind, in saying that
judgment requires an assent of the will.” So, the idea that belief is volitional in a more
general sense is not so far-fetched. Though I recognize that some reserve the term ‘will’
for a narrower sense that is distinctively practical, I’ll continue to invoke the notions of
will and volition in a broader sense to describe the kind of exercise of agency I argue is
involved in believing.
become answerable in the ways Hieronymi claims we do when we commit to the truth of
a proposition. But surely, when I form a belief, such as the belief that it’s 65 degrees
Fahrenheit outside, I don’t directly will that I become answerable in all these ways. So,
how does committing to the truth involve an exercise of will? The answer is that when I
form a belief, I represent p as true by exercising the normative power to subject that very
Think again about what Jacky does when she commits to Joan. In committing to
Joan, understood volitionally, she exercises the normative power to subject herself to
certain normative standards for how she treats Joan. She thus alters the normative
7
landscape, bringing into being normative relations, such as those between Joan’s interests
and her (Jacky’s) reasons for action. This is how she becomes answerable to questions
and criticisms regarding how she treats Joan. These questions and criticisms concern
whether she is living up to the standards she has subjected herself to in committing to
Joan.
Again, we can say something strikingly similar about committing to the truth of p.
true. I thus alter the normative landscape, bringing into being normative relations, such
as those between evidence for the truth of p and my reasons for representing p as true.
This is the sense of commitment I have in mind when I say that to believe p is to commit
to its truth. This isn’t the sense of commitment epistemologists usually invoke. Instead,
surprisingly, I think it’s the very same volitional sense of commitment at play when we
belief consists in a kind of self-binding, in roughly the same sense action does on a
voluntarist interpretation of Kant. In believing, I give the law to myself, where the law is
the normative standard that my belief is correct if and only if it is true. This approach has
some things in common with a range of views that epistemologists call normativism about
belief. But my account is importantly different. Normativists hold that it’s constitutive of
belief that it’s subject to certain norms. For example, Wedgwood (2002, 2007) argues that
8
it’s constitutive of belief that a belief is correct if and only if the proposition believed is
true. 5 For Wedgwood, this normative standard of correctness is just a primitive truth that
that’s precisely what my account does by appealing to our normative power to subject
ourselves to such standards. So, I take things a step further than the normativist. The
normativist tells us that what makes an attitude a belief is that it is subject to this norm
of correctness. But again, while this tells us something important about belief, it doesn’t
tell us anything about what the psychological state of belief is such that it would be this
way. 6 My account discharges this burden and explains why a belief is correct if and only
if the proposition believed is true. It’s because what it is to believe is to exercise our agency
correctness. The normativist is right that it’s constitutive of belief that a belief is correct
if and only if the proposition believed is true. But on my account, this is because it’s more
We now have on the table an answer to the question: in what distinctive sense does
the belief that p involve regarding p as true? The answer is that the belief that p involves
representing p as true by making it the case that that very representation of p as true is
5 See also Engel (2013, 2018), Nolfi (2015), Shah (2003), and Shah and Velleman (2005).
6One exception is what Barranco Lopez (ms) calls commitment normativism about belief. Because she also
sees commitment as a kind of self-binding, her account is closer to mine than it is to what she calls
conventional normativism. But she focuses less than I do on the precise nature of commitment, and more
on the norms generated by our commitments. Another similar view, which Barranco Lopez draws on, is
found in Neta (2018). Neta gives a general account of rationally determinable conditions, including belief,
in terms of commitment. Both he and Barranco Lopez focus on commitment as committing to act, think,
and feel in certain ways. While I agree that commitment generates output-side norms, I discuss in §2.1 why
conceiving of commitment in these terms is insufficient.
9
correct if and only if p is true. Shortly, I’ll show how this commitment-based account of
belief both distinguishes belief from other cognitive attitudes and explains its unique
connection to evidence.
we might see belief as a committed relationship with the truth. It’s part of the nature of
(volitional) commitments that some are exclusive and others are not. For example, it
other friends, subjecting myself to similar or identical normative standards with regard
subject myself to the same normative standards with regard to her as I do with anyone
relationship is exclusive; it rules out making the same commitment to anyone else. We
should view the commitment to the truth that constitutes belief as similarly exclusive. A
standard that it is correct if and only if p is true. This ‘only if’ makes the commitment
This is such an important clarification because it shows why the volitional aspect
of my account doesn’t render it completely absurd. If one could will one’s representation
10
to be subject to other, non-truth-related norms, then belief could be any old thing. But
truth. In other words, belief is not just a committed relationship with the truth, but a
monogamous one too. I’ll build on this point when I discuss how my account explains the
As stated before, the challenge for any account of belief that starts from the idea of
‘regarding as true’ is that there are other attitudes that seem to involve regarding a
proposition as true. Accepting, assuming, and imagining are all plausibly characterized as
regarding-as-true attitudes from belief is to point out that belief involves not just
regarding a proposition as true, but regarding it as true based on evidence of its truth.
We might thus try to distinguish belief from other cognitive attitudes in terms of
this strategy is that it conflates a consequence of the difference between belief and other
regarding-as-true attitudes with the difference itself. To understand what about the
fundamental nature of belief distinguishes it from these other attitudes, we need to know
what in the nature of belief makes it uniquely responsive to evidence. That must be what
argue that this relationship falls out of the fact that belief has truth as its standard of
11
correctness. 7 I will pursue a related strategy. But my explanation goes deeper, because the
as its fundamental normative standard. 8 I’ll now turn to explaining how the commitment-
based account gives us the resources to distinguish belief from other regarding-as-true
attitudes, not just by pointing to distinct features of belief, but by pointing to its distinct
nature.
Accepting, assuming, and imagining that p all seem to involve in some way
regarding p as true. When a lawyer accepts that his client is innocent, he regards that
proposition as true, at least in the context of the courtroom. When I assume there are
objective moral facts for the purposes of some normative ethical argument, I regard the
proposition that there are objective moral facts as true, at least in the context of making
that argument. And when I imagine that I have a million dollars, I regard that proposition
as true within the bounds of the imaginative episode. But belief is different from all of
acceptance in a context, is that the difference between belief and these other regarding-
7 Normativists often explain things in this way. This includes at least Engel (2013), Shah (2003), and Shah
and Velleman (2005). An alternative normativist strategy is that of Nolfi (2018), who argues that evidence-
responsiveness is ultimately explained not by belief’s connection to truth, but instead by the role of
evidence-responsiveness in guiding action.
8 This isn’t to say that normativists never try to explain why beliefs are correct if and only if they are true.
Shah (2003) and Shah and Velleman (2005) argue that it’s because this norm is part of the concept of belief.
So, to conceive of one’s own attitude as a belief requires subjecting it to this correctness norm. But this is
ultimately a thesis about the concept of belief. Even if Shah and Velleman are right about this, I want to
know what it is about the attitude belief itself that would give the concept it picks out this feature.
12
as-true attitudes is that to believe p is to regard p as true independently of any context,
whereas all of these other attitudes involve regarding p as true within some restricted
of one’s deliberation” (2). Acceptance differs from belief partly because “what we
reasonably take for granted, in contrast with what we believe, can vary across different
contexts and be in part shaped by various practical considerations” (4).9 These practical
considerations fix the context within which one takes p for granted. For example, practical
considerations fix that the lawyer takes his client’s innocence for granted only within the
saying something about what belief and acceptance each are that explains why they have
plausible that it involves a commitment to the truth of p in this way. This is because I can
represent p as true by way of taking it for granted in further reasoning, without subjecting
my representation to the standard that it is correct if and only if p is true. The lawyer who
accepts that his client is innocent while in the courtroom doesn’t fail to satisfy a normative
standard if his client is guilty, precisely because he hasn’t bound his representation of his
9Certain defenders of pragmatic encroachment may take issue with this claim, arguing that belief too varies
across different contexts and is in part shaped by practical considerations. This seems like it would make
the distinction between acceptance and belief more difficult to sustain. It would be interesting to explore
whether this is so, and whether it poses a problem for proponents of pragmatic encroachment.
13
Importantly, the volitional conception of commitment to the truth does work here
commitment in mind, one might wonder why this shouldn’t count as commitment to the
truth of p. When one accepts p, one regards p as true in the sense that one commits to a
range of things that seem paradigmatic of treating p as true. Why not think of this as
commitment to the truth of p within the context of the acceptance? If we did, this would
take us back to distinguishing belief from other regarding-as-true attitudes only on the
basis of context-dependence.
of p as subjecting us only to output-side norms – norms that involve acting, thinking, and
difficult to deny that accepting p also involves committing to the truth of p. The only
difference would be that acceptance would involve acting, thinking, and feeling as if p is
subjecting oneself to the input-side norm of representing p as true if and only if p is true.
This is an input-side norm because it regulates how and when we represent p as true,
10 This is part of why I prefer my conception of commitment to Neta’s and Barranco Lopez’s (see fn 5).
Because they understand commitment primarily in terms of output-side norms that involve acting,
thinking, and feeling as if p is true, it seems their views would run into this problem.
14
rather than what happens as a result of our representing p as true. 11 Even if belief and
acceptance are subject to similar output-side norms, with the only difference being
subject one’s representation of p as true to some input-side norm for when it would be
correct to use p as a premise in further reasoning. And thinking about any example of
acceptance within a context, it becomes clear that that correctness norm is going to be a
norm regarding the practical utility of representing p as true in that context, not a truth
norm. Since we have already established that commitment to the truth of p generates an
exclusive input-side truth norm, it follows that the commitment involved in accepting p
can’t be the same commitment as the one involved in believing p. Thus, my account
Here's what we have so far. Belief and acceptance each involve regarding a
proposition as true, but in very different ways. In the case of belief, one represents p as
true by way of willing that one’s representation of p as true is correct if and only if p is
15
true by way of willing that one’s representation of p as true is correct if and only if some
practical conditions are met. This dovetails with Bratman’s contention that unlike belief,
similarly. If I assume the truth of p for the purposes of a proof or argument, I represent p
as true within the context of that proof or argument. But do I commit to the truth of p?
No. I don’t will that my representation of p as true is correct if and only if p is true. If I
did, then anytime I assumed something false for the sake of argument, I would fail to
bound by any standard, it’s some standard involving the fruitfulness of the assumption in
advancing the proof or argument in question. 12 This entails that belief and assumption
can’t be the same thing, for again, they’re constituted by different commitments.
Imagining that p is an interesting case, because it isn’t clear that imagination has
any constitutive normative standards. While I might imagine certain things for certain
purposes, I could also just idly imagine different scenarios while letting my mind wander.
In such a case, while I might be representing what I imagine as true within the context of
the imaginative episode, it doesn’t seem like this representation is bound by any
normative standards at all. I can simply let my imagination wander. Once again, we have
a principled difference between the nature of belief and imagination. While belief is
12 This is why, when I assume (incorrectly), I make an ass out of you and me.
16
constituted by a commitment to the truth of a proposition, imagining doesn’t seem to be
One of the most important ways in which belief is distinct from other regarding-
as-true attitudes is that there seems to be a uniquely tight connection between belief and
the believer takes as evidence. On the normative side, it seems like there’s a corresponding
evidential norm roughly of the form that the believer should hold a belief only on the basis
the basis of considerations that I take to be irrelevant to its truth. In fact, these attitudes
Nor have I made any normative mistake if I accept, assume, or imagine on the basis of
practical considerations.
13 On a related note, we might wonder how the commitment-based account of belief handles ‘edge cases’
like delusions or implicit biases. Both delusions and implicit biases plausibly involve representations on the
part of the agent. However, they will only count as beliefs if they are constituted by a commitment to the
truth of a proposition. Whether or not it’s plausible to construe delusions and implicit biases in this way
will depend on how we understand these phenomena. For example, if we understand them as kinds of
recalcitrant representations that are contrary to an agent’s commitments, then they won’t count as beliefs.
Thinking about mental states we might be alienated from on my framework raises interesting questions
about the possibility of being alienated from our own wills, which I don’t have space to address in this essay.
14 Another attitude distinct from belief that might involve regarding as true is perception, at least on some
propositional accounts. My account also provides the resources to distinguish belief from perception, so
understood, because perception does not plausibly involve commitment. Thanks to an anonymous reviewer
for suggesting I consider this.
17
The commitment-based account of belief is well-placed to explain the relationship
between belief and evidence. First, consider the normative side. On my account, when one
believes some proposition p, one subjects one’s representation of p as true to the norm
that it is correct if and only if p is true. From here, the commitment-based account can
give a similar story to the one given by normativists. On a standard normativist account,
evidential norms on belief are derivative from the fundamental norm that the belief that
The basic idea is this: evidence for p is, ipso facto, evidence for the truth of p. Why?
Because evidence is the subjective shadow of objective truth. Given our epistemic
limitedness, we can only follow the fundamental truth norm of belief by looking for
indicators of truth – pieces of evidence. 15 As Wedgwood puts it, the point of conforming
to evidential standards of rational or justified belief is “to ensure that one believes the
proposition in question if and only if that proposition is true” (2007, 154). The normative
connection between belief and evidence in this way depends on the normative connection
The commitment-based account of belief explains how the standard of truth for
our representations to this standard. But I haven’t said anything about subjecting our
15 It's far beyond the scope of this essay to give an account of what evidence for a proposition is, besides
something that indicates the truth of it. One way of looking at it is that evidence for a proposition
probabilifies its truth – the probability of the proposition is higher given that piece of evidence than without
it. Whatever the precise relation is, pretty much everyone agrees that evidence for p is a guide to or indicator
of the truth of p. So, it’s very natural to think of a norm of sufficient evidence as derivative from a norm of
truth.
18
do evidential standards get to be normative? Consider again the analogy between
commit to another person, I make myself answerable not just to compliance with the
normative standard I directly impose on myself, but also to compliance with a range of
derivative standards. It would be patently absurd if Jacky were to say to Joan, “well, I
committed to you, but I never committed to doing x, y, and z,” where doing x, y, and z is
Thus, we can conclude that exercising the normative power to subject oneself to
some normative standard also makes one answerable for complying with whatever
standards are derivative, in the sense of being necessary for compliance with the
fundamental standard. Given our epistemic limitations – i.e., our lack of direct, infallible
access to the truth with regard to most beliefs – complying with evidential norms is the
only way to attempt to comply with the fundamental truth norm of belief. This is why,
though we don’t directly will ourselves to be subject to evidential norms, those standards
Now to the psychological side of the relationship between belief and evidence. As
responsive to what we take as evidence. Even though this connection is imperfect, it still
distinguishes belief from other regarding-as-true attitudes that lack it. But how do we
16An interesting implication of this is that if there are beliefs such that it’s not constitutive of complying
with the truth norm that we comply with the norm of sufficient evidence, we wouldn’t be subject to the latter
norm in such cases. This seems to me like the correct result. Indeed, it allows us to account for the tight
normative connection between belief and evidence while allowing that certain kinds of beliefs (e.g., basic
beliefs) are not subject to evidential norms. This may help deal with Barranco Lopez’s (ms) pressing
objection to input-side evidential norms that they can’t apply to hinge-proposition beliefs.
19
explain this connection? It won’t do to merely point to the normative authority of belief’s
standard of correctness. After all, there are lots of genuine norms we seem to have little
The most obvious answer is that something in the nature of belief must explain
why this norm has not just normative but also psychological purchase on us. For Shah
and Velleman, it’s that the standard of correctness is part of the concept of belief such that
“one is forced to apply the standard of correctness…in situations in which one exercises
the concept of belief” (2005, 501). Shah (2006) argues that because deliberation about
what to believe exercises the concept of belief, it necessarily involves applying the
standard of correctness. In asking the question whether to believe p, one recognizes that
one’s belief is correct if and only if p is true. This in turn “sets the standard for what can
count as a reason for or against believing that p” (493). So, insofar as one is answering the
question whether to believe p, one must answer it by bringing to bear considerations one
belief can better explain the psychological side of evidence-responsiveness. For Shah and
Velleman, the standard of correctness is part of the concept of belief, so it binds our
representations only insofar as we deploy the concept of belief. Moreover, even if I deploy
the concept of belief and thereby recognize what can and can’t count as a reason for
believing p, why can’t I just flout this requirement? The answer must be that if I flout it,
my attitude doesn’t count as a belief. But it doesn’t seem like anything about the concept
of belief could get us that result; at best, it could get us the result that if I flout this
20
On my account, by contrast, belief’s standard of correctness is binding because in
standard, I place my agency behind that standard in a way that makes it impossible for
me to self-consciously flout it. I might fail to comply with it, just like we might generally
fail to live up to our commitments. But insofar as I am committed to the truth of p, it isn’t
this commitment, I similarly must be responsive to what I take as evidence. If I were not,
the resulting commitment wouldn’t be a sincere commitment to the truth. This gets us
exactly what we need: that belief has a tight but imperfect psychological connection to
evidence. I can fail to live up to my commitment to the truth by, say, engaging in wishful
thinking or other cognitive biases. But I can’t, while retaining this commitment, flout it
its truth.
The above discussion should also assuage any lingering worries from evidentialists
and other epistemic purists that the involvement of the will in my account of belief is a
norms it gives rise to, the commitment-based account of belief seems to push us toward
any attitude I formed on that basis would fail to be a belief. This is exactly the kind of tight
21
Indeed, my account is particularly congenial to arguments by Kelly (2002) and
considerations to the truth of evidentialism. The basic idea behind this form of argument
is that, if it’s impossible to believe on the basis of some type of consideration, that
consideration can’t constitute a normative reason for belief. Since we can only believe on
normative reasons for belief (i.e. evidentialism is true). Insofar as the commitment-based
account explains the impossibility premise in such arguments, it provides even more
So far, I have provided a case for the commitment-based account by showing how
it distinguishes belief from other regarding-as-true attitudes and explains belief’s unique
connection to evidence. One of the greatest virtues of the account, I think, is that it shows
how all of this falls out of the fundamental nature of belief itself, rather than the concept
of belief, or some derivative feature of belief. But it has another significant virtue, which
is that it helps solve several puzzles about belief. I’ll discuss three such puzzles in this
essay: doxastic voluntarism and epistemic deontology; the aim of belief; and Moore’s
paradox.
17Crucially, though, the commitment-based account only delivers the result that it’s impossible to believe p
on the basis of considerations that one takes to be irrelevant to the truth of p. This can come apart from
what actually is or isn’t evidence. Worsnip (2021) argues that because of this gap, the Kelly-Shah argument
for evidentialism falls apart. I’m optimistic that it can be salvaged, but I don’t have space to argue for that
here.
22
3.1. Doxastic voluntarism and epistemic deontology
subject one’s representation of p as true to the normative standard that it is correct if and
only if p is true. This means belief involves an exercise of volition, something that much
hold that belief is, in contrast to action, paradigmatically non-voluntary. Action and
practical reasoning involve the will, but belief and theoretical reasoning cannot. That’s
As Williams illustrates, the problem with the idea of believing at will is that it would mean
being able to decide to believe anything at all, regardless of its connection to truth. And
Doxastic voluntarism is the thesis that agents can form beliefs voluntarily. Because
voluntarism is generally thought to be false. This might be seen as a problem for a view
like mine, on which belief is voluntary. However, the sense in which belief is voluntary on
my account is different from the sense Williams had in mind. On my account, though
belief involves volition, one can’t believe by willing to acquire a belief. Rather, one’s
itself constituted by an exercise of the normative power to subject that very representation
to the normative standard of truth. Thus, the will in belief involves not a will to believe,
23
but a will to truth. Far from implying that we can acquire beliefs irrespective of their truth,
objectionable sense.
be done arbitrarily, or willy-nilly. There are certain kinds of activities that can be
voluntary, and yet if we don’t do them in a certain way, it doesn’t count as an instance of
that activity. For example, I doubt anyone would argue that sincere assertions are
involuntary. But sincere assertions are like beliefs in precisely the way Williams describes:
reasonable to question whether sincere assertion was really what I had in mind. The fact
that I can only sincerely assert p if I am committed to the truth of p does not suggest that
sincerely asserting p is involuntary. Similarly, the fact that I can only believe p if I am
But perhaps the problem with doxastic voluntarism isn’t really that it allows for
‘believing at will.’ Perhaps the problem is that it is supposed to allow for believing
something irrespective of its truth in the sense of believing on the basis of truth-irrelevant
practical reasons. For example, Bennett argues that beliefs are not voluntary because
“take φing to be voluntary if one's φing depends upon inducements, that is, if one can φ
or not depending on whether one thinks one has practical reasons to φ” (1990, 90).18 But
again, this is not the sense in which belief is voluntary on my account. I mean voluntary
18Hieronymi (2006) rejects doxastic voluntarism based on Bennett’s understanding of voluntariness but
argues that we nonetheless have evaluative control over our beliefs. My disagreement with Hieronymi (and
Bennett) here is more about how to understand voluntariness than it is the nature of belief. See also Shah
(2002) for important commentary on different ways of understanding voluntariness and voluntarism.
24
in the simple sense of involving volition, understood broadly as the agential power to
based on an identification of volition with practical reason that I have already addressed
in §1.2.
Now that I’ve clarified the sense in which belief is voluntary, I can explain how my
account solves the puzzle of doxastic voluntarism and epistemic deontology. The puzzle
stems from an argument originally made by Alston (1988), which goes roughly like this:
This argument has been taken to generate a puzzle because the conclusion denies
something that seems quite plausible. It makes good sense to talk about what we ought or
ought not believe, among other forms of deontic evaluation. But each of the two premises
The problem with the argument is that premise 2 is most plausible if we interpret
sense. But if we interpret doxastic voluntarism in that way, then premise 1 is false. It isn’t
necessary to be able to believe at will for our beliefs to be subject to deontic evaluation.
That only seems necessary if we erroneously identify the lack of ability to believe at will
with belief’s being a merely passive state over which we have no control. Of course, if we
have no control over our beliefs, then deontic evaluation seems inappropriate. But as both
Shah (2002) and Hieronymi (2006) persuasively argue, despite not believing ‘at will,’ we
25
control our beliefs by forming them in response to the right kind of reasons. Moreover,
this is the very same kind of control we have over our intentions.
Indeed, Shah argues that it’s precisely in this sense that doxastic voluntarism is
true, and thus rejects Bennett’s narrow conception of voluntariness. I agree, and I think
the commitment-based account of belief explains why we should think of this kind of
exercise of volition. The sense in which belief is voluntary is that it involves a commitment
to the truth that is undertaken voluntarily, but where only that commitment could
constitute that belief. Thus, if we accept my account, the puzzle of doxastic voluntarism
But this is a puzzling claim to try to unpack. The first instance I know of a philosopher
claiming that “beliefs aim at truth” is Williams (1973, 136-137). Williams writes he has in
mind three things by this: (1) that truth is distinctively a dimension of normative
assessment for beliefs, (2) that to believe p is to believe that p is true, and (3) to assert “I
believe that p” is to claim that p is true. But these are three different claims, and “beliefs
aim at truth” can’t mean all three of them. Indeed, Williams writes further that these three
claims are “vaguely summed up” in the slogan that beliefs aim at truth (173). So for
19 The arguments in this section build heavily on material from Singh (2022).
26
Many have taken the idea that belief aims at the truth to be more than just a slogan
literally have the aim of believing the truth: “believing involves regarding a proposition as
true with the aim of so regarding it only if it really is. Thus, to believe a proposition is to
accept it with the aim of thereby accepting a truth” (2000, 251). For the early Velleman,
this is an aim in the teleological sense, one that could be realized by the believer’s
executing an intention to accept the truth in regarding a proposition as true. This aim
could also be realized more implicitly by some kind of causal regulation of the believer’s
cognitive systems for truth, which Velleman argued still counts as attempting to accept
the truth.
This broadly teleological conception of the aim of belief has the advantage of
making it more than just a vague metaphor. However, it’s been widely rejected, including
by Velleman himself, on account of what Shah (2003) calls the teleologist’s dilemma.
Briefly, the teleologist’s dilemma arises when we try to explain why truth-related
considerations play an exclusive role in doxastic deliberation by appealing to the fact that
belief aims at the truth. The teleologist must take the aim of belief to consist in belief’s
being either weakly or strongly causally regulated for truth. If belief is weakly causally
regulated for truth, then it is possible for non-truth-related considerations to play a role
in causally regulating our beliefs. So, truth-related considerations would fail to play an
exclusive role in doxastic deliberation. If belief is strongly regulated for truth, this yields
the result that truth-related considerations play an exclusive role in doxastic deliberation,
but at the cost of entailing that wishful thinking and other cognitive biases are impossible.
27
Partly on this basis, teleologism has largely been rejected in favor of normativism
about the aim of belief. 20 According to normativists, the idea that belief aims at the truth
is ultimately a metaphor after all, for the normativist thesis that belief has the constitutive
norm of being correct if and only if the proposition believed is true. As Wedgwood puts it,
“Beliefs are not little archers armed with little bows and arrows: they do not literally ‘aim’
at anything” (2002, 267). So, talk of constitutive aims is just an evocative, metaphorical
This way of thinking about the aim of belief renders the ‘aim’ language puzzling,
though. Why would we use language that so much suggests we’re talking about something
psychological, if the aim of belief really refers to something purely normative? At least for
teleologists like the early Velleman, the aim of belief is something about what it is,
psychologically speaking, to hold the attitude of belief. To be fair to Shah and the later
Velleman, they do connect the aim of belief to psychology, with the arguments I discussed
in §2.2 about the correctness norm being a conceptual truth, and the necessity of
deploying the concept of belief in doxastic deliberation. But I find this connection
unsatisfyingly indirect, for it only goes by way of our application of the concept of belief.
I think the commitment-based account of belief can provide a good middle ground
between teleologism and normativism, one that avoids the teleologist’s dilemma while
preserving the idea that the aim of belief picks out a psychological fact about the truth-
directed nature of the attitude. On my account, the sense in which belief aims at the truth
20 See Shah (2003), Shah and Velleman (2005), and Wedgwood (2002).
28
standard of correctness in virtue of the psychological fact of one’s willing that it be so. So,
in believing, we direct our representations toward the truth in a very literal sense. This
bridges the gap between the teleologist’s psychological interpretation and the
account provides a more satisfying explanation of what it means for belief qua
deliberation, without entailing that cognitive bias is impossible. Indeed, I have already
provided this explanation in §2.2. If my beliefs are influenced by cognitive bias, I fail to
fully live up to my commitment to the truth. This is compatible with my retaining that
of considerations I take to be irrelevant to its truth is not compatible with retaining that
Our sincere assertions express our beliefs. If I sincerely assert both p and ~p, I’m
expressing both a belief that p and a belief that ~p. It’s fairly straightforward to explain
what seems off about this. In asserting a contradiction, I express two incompatible stances
on whether p is true. Moore-paradoxical assertions, by contrast, are of the form “p, but I
don’t believe p” (omissive) or “p, but I believe ~p” (commissive). Because it could be true
that p, but also true that I don’t believe p (or believe ~p), Moore-paradoxical assertions
don’t seem obviously contradictory in the way asserting both p and ~p is. The two
29
belief about whether p is true, and a second-order belief about what I believe about p.
Because we can believe falsehoods, and fail to believe truths, what we believe and what is
the case can obviously come apart. So why do Moore-paradoxical assertions seem just as
One type of solution invokes the claim that believe is self-intimating – if I believe
p, then I believe that I believe p. So, if by asserting p, I express the belief that p, the belief
I directly express intimates that I believe that I believe p, then, by asserting the other
conjunct “I don’t believe that p,” I express the belief that I don’t believe p, thereby
contradicting something intimated by the first conjunct. On this way of accounting for the
between the conjuncts of the assertion. Thus, whereas asserting p and ~p would be an
The problem with this solution, however, is that this seems to undersell the
asserting first-order contradictions. What the inter-level conflict solution misses is that,
seem to express incompatible stances on whether p is true. But if the inter-level conflict
solution is right, what we get instead, through the ascent from the expressed belief that p
to the intimated belief that I believe that p, is incompatible stances on whether I believe
Unlike the inter-level conflict solution, my account of belief provides the resources
30
whether p is true. Recall the quotation from Alston earlier that “in making statements and
holding beliefs we commit ourselves to the propositional content's being true.” I think
Alston is right that assertions too (at least, sincere ones) involve commitments. Here’s
how that works: the relation between sincere assertion and belief is that, in asserting p, I
express the very same commitment to the truth of p that I make in believing p. This is why
Now, take the second conjunct of the Moore-paradoxical assertion. What happens
when I assert that I don’t believe p? The surface grammar of the assertion may suggest
that I express a second-order belief about what I believe, but this is misleading. Because
believe p express whether or not I am committed to the truth of p. 21 So, rather than
commitments. This illuminates the solution to Moore’s paradox: each conjunct of the
regard to the truth of p. When one asserts, “p, but I don’t believe p,” one at once expresses
a commitment to the truth of p and a lack thereof. When one asserts, “p, but I believe ~p,”
one at once expresses a commitment to the truth of p and a commitment to the truth of
~p.
whether p is true – in one case, both commitment and a lack thereof, and in the other,
21 This is why, as Williams (1973, 137) writes, “to say ‘I believe that p’ itself carries, in general, a claim that
p is true.” Indeed, Williams explicitly connects this feature of belief to the absurdity of Moore-paradoxical
assertions.
31
opposing commitments. The conflict is wholly first-order. Indeed, on my account, the
commissive version of the assertion expresses the same kind of contradiction expressed
by asserting both p and ~p. In this way, the commitment-based account provides a
superior solution to Moore’s paradox, one that makes clear how Moore-paradoxical
Concluding Remarks
There’s much more to say about belief as commitment to the truth than can fit in
one essay. But I hope to have at least made a convincing case for the explanatory power
the truth of p. I understand the relevant kind of commitment volitionally, as the exercise
standard that it is correct if and only if p is true. I’ve argued that this account has great
attitudes, such as acceptance, assumption, and imagination. It also has great explanatory
power when it comes to both the psychological and normative aspects of the relationship
between belief and evidence. Finally, it provides the resources to solve a variety of puzzles
about belief, including doxastic voluntarism and epistemic deontology, the aim of belief,
22 Marcus (2021. ch. 3) defends a similar solution to Moore’s paradox, but his doesn’t invoke the notion of
commitment. On Marcus’ account, in uttering “p but I don’t believe that p,” one simultaneously asserts that
p and disavows the belief that p. This expresses two incompatible stances on the question whether p: belief
and abstention. While I agree with much of what Marcus argues, his solution relies on certain claims about
self-knowledge that mine doesn’t.
23 Woods (2018) defends a commitment-based solution to Moore’s paradox. However, Woods has a very
different sense of commitment in mind from mine. For Woods, commitments are not volitional or even
psychological, but rather social facts about what we ought to do given particular systems of formal norms.
32
Despite all of this, I’m sure there will be strong opposition to my account of belief.
The idea that belief involves volition may strike some as just too unorthodox to swallow.
But I think it speaks in favor of it that a seemingly heterodox account of belief can account
for so many of the features orthodox epistemology takes belief to have. Nevertheless, I
wish I had more space to address the many pressing objections to the account developed
here. I will conclude by briefly considering just one, which is perhaps the most obvious
I’ve claimed that each belief that p consists in a representation of p as true that
subjects that very representation to the normative standard that it is correct if and only if
p is true. We have a lot of beliefs. Isn’t this a bit much to claim is going on in our minds
implausibly messy and cluttered picture of our psychology? It depends. The commitment-
based account applies most plausibly and straightforwardly to our explicit, avowed beliefs
– the ones we’ve at some point asserted, out loud or in our heads, and really thrown our
agency behind. How far the account can be built out from there is a difficult question.
But here’s a brief thought about how this might work. Aside from our explicit
beliefs, we have lots of implicit, or tacit, beliefs. 24,25 Representationalists generally hold
that only explicit beliefs are stored in the mind. Implicit beliefs are those that are swiftly
derivable from our explicit beliefs, but which we’ve never specifically thought about
24 On this topic, see Harman (1986) and Lycan (1986), among others.
25 Thanks to Andrew Moon for suggesting that I address the subject of implicit/tacit beliefs.
33
before, like the belief that I’m less than 143 inches tall. 26 Does the commitment-based
view have anything interesting to say about this, over and above the standard
representationalist line?
Perhaps it does. Harman (1986) uses the term “implicit commitments” to refer to
things we don’t explicitly believe but would be incoherent for us to deny given what we
explicitly believe. Though Harman doesn’t dwell on the notion of commitment, perhaps
if an implicit belief is a belief that would be easily derivable from what we explicitly
what we are explicitly committed to, insofar as we live up to those explicit commitments.
This would fit with the commitment-based account. Moreover, it might shed light on the
things we’re rationally committed to, like the easily derivable logical consequences of our
explicit beliefs, are just the things we’re implicitly volitionally committed to, in the sense
that we would easily derive volitional commitment to them from our explicit volitional
26For an objection to the standard representationalist line on implicit/tacit beliefs, see Schwitzgebel (this
volume).
34
References
Bratman, Michael E. (1992). Practical reasoning and acceptance in a context. Mind 101
(401):1-16.
Chang, R. (2009). Voluntarist reasons and the sources of normativity. In D. Sobel & S.
Wall (Eds.), Reasons for Action (pp. 243–271). Cambridge University Press.
Engel, P. (2013). In Defence of Normativism about the Aim of Belief. In T. Chan (Ed.),
The Aim of Belief (pp. 32–63). Oxford University Press.
Engel, Pascal (2018). The Doxastic Zoo. In Annalisa Coliva, Paolo Leonardi &
Sebastiano Moruzzi (eds.), Eva Picardi on Language, Analysis and History.
Palgrave. pp. 297-316.
Hieronymi, P. (2005). The Wrong Kind of Reason: Journal of Philosophy, 102(9), 437–
457.
35
Hieronymi, Pamela (2009). The will as reason. Philosophical Perspectives 23 (1):201-
220.
Kelly, Thomas (2002). The rationality of belief and other propositional attitudes.
Philosophical Studies 110 (2):163-96.
Lycan, William G. (1986). Tacit belief. In R. Bogdan (ed.), Belief: Form, Content, and
Function. Oxford University Press.
Marcus, Eric (2021). Belief, Inference, and the Self-Conscious Mind. Oxford, United
Kingdom: Oxford University Press.
Nolfi, Kate (2015). How to be a Normativist about the Nature of Belief. Pacific
Philosophical Quarterly 96 (2):181-204.
Nolfi, Kate (2018). Why Only Evidential Considerations Can Justify Belief, in Conor
McHugh, Jonathan Way, and Daniel Whiting (eds), Normativity: Epistemic and
Practical. Oxford University Press.
Shah, Nishi (2002). Clearing Space for Doxastic Voluntarism. The Monist 85 (3):436-
445.
Shah, N. (2003). How Truth Governs Belief. Philosophical Review, 112(4), 447–482.
Shah, N., & Velleman, J. D. (2005). Doxastic Deliberation. The Philosophical Review,
114(4), 497–534.
36
Wedgwood, Ralph (2007). The Nature of Normativity. Oxford, GB: Oxford University
Press.
37